Lucene 时间排序
在Lucene4.4中,想要实现搜索结果按照时间倒序的效果:如果两个文档得分相同,那么就按照发布时间倒序排列;否则就按照分数排列。这种效果在Lucene4.6中实现起来极其简单,直接利用search接口的Sort参数即可达成,完全不需要像某些人说的重写Similarity那么麻烦。三两行代码的事情,体现了Make it simple, stupid的精髓。
首先来看看测试例子,这个例子中我建立了四个文档,按照内容-发布日期来表示分别是:
2004年光棍节攻略 , 20041111
2005年光棍节攻略 , 20051111
2006年光棍节攻略 , 20061111
游戏攻略 ,20141111
统一使用“光棍节攻略”来搜索它们,用户希望最新的光棍节攻略排在第一。
如果不做排序处理的话,用户体验非常糟糕:
- package com.hankcs.test;
- import org.apache.lucene.analysis.Analyzer;
- import org.apache.lucene.document.*;
- import org.apache.lucene.index.*;
- import org.apache.lucene.queries.CustomScoreQuery;
- import org.apache.lucene.queries.function.FunctionQuery;
- import org.apache.lucene.queryparser.classic.ParseException;
- import org.apache.lucene.queryparser.classic.QueryParser;
- import org.apache.lucene.search.*;
- import org.apache.lucene.store.Directory;
- import org.apache.lucene.store.LockObtainFailedException;
- import org.apache.lucene.store.RAMDirectory;
- import org.apache.lucene.util.Version;
- import org.wltea.analyzer.lucene.IKAnalyzer;
- import java.io.IOException;
- /**
- * @author hankcs
- */
- public class TestSortByTime
- {
- public static void main(String[] args)
- {
- // Lucene Document的主要域名
- String fieldName = "text";
- // 实例化IKAnalyzer分词器
- Analyzer analyzer = new IKAnalyzer();
- Directory directory = null;
- IndexWriter iwriter;
- IndexReader ireader = null;
- IndexSearcher isearcher;
- try
- {
- //索引过程**********************************
- //建立内存索引对象
- directory = new RAMDirectory();
- //配置IndexWriterConfig
- IndexWriterConfig iwConfig = new IndexWriterConfig(Version.LUCENE_46, analyzer);
- iwConfig.setOpenMode(IndexWriterConfig.OpenMode.CREATE_OR_APPEND);
- iwriter = new IndexWriter(directory, iwConfig);
- //写入索引
- for (int i = 0; i < 3; ++i)
- {
- int year = 2004 + i;
- Document doc = new Document();
- doc.add(new TextField(fieldName, year + "年光棍节攻略", Field.Store.YES));
- doc.add(new IntField("date", year * 10000 + 1111, Field.Store.YES));
- iwriter.addDocument(doc);
- }
- // 加入一个干扰文档
- Document doc = new Document();
- doc.add(new TextField(fieldName, "游戏攻略", Field.Store.YES));
- doc.add(new IntField("date", 20141111, Field.Store.YES));
- iwriter.addDocument(doc);
- iwriter.close();
- //搜索过程**********************************
- //实例化搜索器
- ireader = DirectoryReader.open(directory);
- isearcher = new IndexSearcher(ireader);
- String keyword = "光棍节攻略";
- //使用QueryParser查询分析器构造Query对象
- QueryParser qp = new QueryParser(Version.LUCENE_46, fieldName, analyzer);
- Query query = qp.parse(keyword);
- System.out.println("Query = " + query);
- //搜索相似度最高的5条记录
- TopDocs topDocs = isearcher.search(query, 5);
- System.out.println("命中:" + topDocs.totalHits);
- //输出结果
- ScoreDoc[] scoreDocs = topDocs.scoreDocs;
- for (int i = 0; i < Math.min(5, scoreDocs.length); i++)
- {
- Document targetDoc = isearcher.doc(scoreDocs[i].doc);
- System.out.print(targetDoc.getField(fieldName).stringValue());
- System.out.print(" , " + targetDoc.getField("date").numericValue());
- System.out.println(" , " + scoreDocs[i].score);
- }
- } catch (CorruptIndexException e)
- {
- e.printStackTrace();
- } catch (LockObtainFailedException e)
- {
- e.printStackTrace();
- } catch (IOException e)
- {
- e.printStackTrace();
- } catch (ParseException e)
- {
- e.printStackTrace();
- } finally
- {
- if (ireader != null)
- {
- try
- {
- ireader.close();
- } catch (IOException e)
- {
- e.printStackTrace();
- }
- }
- if (directory != null)
- {
- try
- {
- directory.close();
- } catch (IOException e)
- {
- e.printStackTrace();
- }
- }
- }
- }
- }
输出:
2004年光棍节攻略 , 20041111 , 0.71185887
2005年光棍节攻略 , 20051111 , 0.71185887
2006年光棍节攻略 , 20061111 , 0.71185887
游戏攻略 , 20141111 , 0.049675122
可以看到文档是严格按照分数排序的,如果分数相同,则按照索引顺序排序,导致最新的文章反而排在最下面。
使用search接口的Sort参数优化搜索结果:
- package com.hankcs.test;
- import org.apache.lucene.analysis.Analyzer;
- import org.apache.lucene.document.*;
- import org.apache.lucene.index.*;
- import org.apache.lucene.queries.CustomScoreQuery;
- import org.apache.lucene.queries.function.FunctionQuery;
- import org.apache.lucene.queryparser.classic.ParseException;
- import org.apache.lucene.queryparser.classic.QueryParser;
- import org.apache.lucene.search.*;
- import org.apache.lucene.store.Directory;
- import org.apache.lucene.store.LockObtainFailedException;
- import org.apache.lucene.store.RAMDirectory;
- import org.apache.lucene.util.Version;
- import org.wltea.analyzer.lucene.IKAnalyzer;
- import java.io.IOException;
- /**
- * @author hankcs
- */
- public class TestSortByTime
- {
- public static void main(String[] args)
- {
- // Lucene Document的主要域名
- String fieldName = "text";
- // 实例化IKAnalyzer分词器
- Analyzer analyzer = new IKAnalyzer();
- Directory directory = null;
- IndexWriter iwriter;
- IndexReader ireader = null;
- IndexSearcher isearcher;
- try
- {
- //索引过程**********************************
- //建立内存索引对象
- directory = new RAMDirectory();
- //配置IndexWriterConfig
- IndexWriterConfig iwConfig = new IndexWriterConfig(Version.LUCENE_46, analyzer);
- iwConfig.setOpenMode(IndexWriterConfig.OpenMode.CREATE_OR_APPEND);
- iwriter = new IndexWriter(directory, iwConfig);
- //写入索引
- for (int i = 0; i < 3; ++i)
- {
- int year = 2004 + i;
- Document doc = new Document();
- doc.add(new TextField(fieldName, year + "年光棍节攻略", Field.Store.YES));
- doc.add(new IntField("date", year * 10000 + 1111, Field.Store.YES));
- iwriter.addDocument(doc);
- }
- // 加入一个干扰文档
- Document doc = new Document();
- doc.add(new TextField(fieldName, "游戏攻略", Field.Store.YES));
- doc.add(new IntField("date", 20141111, Field.Store.YES));
- iwriter.addDocument(doc);
- iwriter.close();
- //搜索过程**********************************
- //实例化搜索器
- ireader = DirectoryReader.open(directory);
- isearcher = new IndexSearcher(ireader);
- String keyword = "光棍节攻略";
- //使用QueryParser查询分析器构造Query对象
- QueryParser qp = new QueryParser(Version.LUCENE_46, fieldName, analyzer);
- Query query = qp.parse(keyword);
- System.out.println("Query = " + query);
- //搜索相似度最高的5条记录
- Sort sort = new Sort(new SortField("text", SortField.Type.SCORE), new SortField("date", SortField.Type.INT, true));
- TopDocs topDocs = isearcher.search(query, 5, sort);
- System.out.println("命中:" + topDocs.totalHits);
- //输出结果
- ScoreDoc[] scoreDocs = topDocs.scoreDocs;
- for (int i = 0; i < Math.min(5, scoreDocs.length); i++)
- {
- Document targetDoc = isearcher.doc(scoreDocs[i].doc);
- System.out.print(targetDoc.getField(fieldName).stringValue());
- System.out.print(" , " + targetDoc.getField("date").numericValue());
- System.out.println(" , " + scoreDocs[i].score);
- }
- } catch (CorruptIndexException e)
- {
- e.printStackTrace();
- } catch (LockObtainFailedException e)
- {
- e.printStackTrace();
- } catch (IOException e)
- {
- e.printStackTrace();
- } catch (ParseException e)
- {
- e.printStackTrace();
- } finally
- {
- if (ireader != null)
- {
- try
- {
- ireader.close();
- } catch (IOException e)
- {
- e.printStackTrace();
- }
- }
- if (directory != null)
- {
- try
- {
- directory.close();
- } catch (IOException e)
- {
- e.printStackTrace();
- }
- }
- }
- }
- }
输出结果:
命中:4
2006年光棍节攻略 , 20061111 , NaN
2005年光棍节攻略 , 20051111 , NaN
2004年光棍节攻略 , 20041111 , NaN
游戏攻略 , 20141111 , NaN
我们看到“2006年光棍节攻略”因为时间比较新,并且相关性高,就排在了第一。“2005年光棍节攻略”相关度相同,因为时间旧就排在后面一点,而干扰文档“游戏攻略”即使时间最新,因为不相关的原因排在最后面。这种效果正好是我想要的,极大提升了用户体验。